bug#65622: Inappropriate suppression of backtrace on an error

bug#65622: Inappropriate suppression of backtrace on an error

[Alan Mackenzie]

Hello, Emacs.

On a recent master branch Emacs:
(i) emacs -Q
(ii) Insert the following into *scratch*:

(defmacro hash-if (condition then-form &rest else-forms)
  "A conditional compilation macro analogous to C's #if.
Evaluate CONDITION at macro-expansion time.  If it is non-nil,
expand the macro to THEN-FORM.  Otherwise expand it to ELSE-FORMS
enclosed in a `progn' form.  ELSE-FORMS may be empty."
  (declare (indent 2)
           (debug (form sexp &rest sexp)))
  (if (eval condition lexical-binding)
      then-form
    (cons 'progn else-forms)))

(defun foo (bar)
  (hash-if (< emacs-major-version 19)
      (car bar)
    (cons bar bar)))

(iii) Evaluate hash-if by putting point after it and doing C-x C-e.
(iv) Attempt to instrument foo for edebug by putting point inside foo and
  doing C-u C-M-x.  This throws the error: "Ignoring macroexpansion
  error: (void-function edebug-after)".
(v) Set debug-on-error to t with M-: (setq debug-on-error t).
(vi) Repeat (iv).  This throws the same error, without a backtrace.  This
  lack of a backtrace is a bug.
(vii) This backtrace is almost certainly being suppressed by a frivolous
  condition-case, whose main purpose appears to be making debugging more
  difficult.  ;-)
(viii) There would appear to be no justification for "ignoring" the error
  (void-function edebug-after).  Such error should not occur, and needs
  debugging.

-- 
Alan Mackenzie (Nuremberg, Germany).

bug#65622: Inappropriate suppression of backtrace on an error

[Michael Heerdegen]

Alan Mackenzie <acm@muc.de> writes:

> Hello, Emacs.
> (v) Set debug-on-error to t with M-: (setq debug-on-error t).
> (vi) Repeat (iv).  This throws the same error, without a backtrace.  This
>   lack of a backtrace is a bug.

I think you just need (setq eval-expression-debug-on-error t).

Stumbled over the same issue ... yesterday?, or so.  Also when debugging
Edebug.

Maybe people would like debug-on-error -> t taking precedence over
eval-expression-debug-on-error -> nil?  The danger is they turn both off
by default and then forget about the second one.

Michael.

bug#65622: Inappropriate suppression of backtrace on an error

[Gerd Möllmann]

Michael Heerdegen <michael_heerdegen@web.de> writes:

> Maybe people would like debug-on-error -> t taking precedence over
> eval-expression-debug-on-error -> nil?  

FWIW, I would.

bug#65622: Inappropriate suppression of backtrace on an error

[Eli Zaretskii]

> Cc: Alan Mackenzie <acm@muc.de>, 65622@debbugs.gnu.org
> From: Gerd Möllmann <gerd.moellmann@gmail.com>
> Date: Thu, 31 Aug 2023 09:41:22 +0200
> 
> Michael Heerdegen <michael_heerdegen@web.de> writes:
> 
> > Maybe people would like debug-on-error -> t taking precedence over
> > eval-expression-debug-on-error -> nil?  
> 
> FWIW, I would.

That'd probably require a special non-nil, non-t value of
debug-on-error, to avoid surprising others.

bug#65622: Inappropriate suppression of backtrace on an error

[Gerd Möllmann]

On 31.08.23 09:49, Eli Zaretskii wrote:
>> Cc: Alan Mackenzie <acm@muc.de>, 65622@debbugs.gnu.org
>> From: Gerd Möllmann <gerd.moellmann@gmail.com>
>> Date: Thu, 31 Aug 2023 09:41:22 +0200
>>
>> Michael Heerdegen <michael_heerdegen@web.de> writes:
>>
>>> Maybe people would like debug-on-error -> t taking precedence over
>>> eval-expression-debug-on-error -> nil?
>>
>> FWIW, I would.
> 
> That'd probably require a special non-nil, non-t value of
> debug-on-error, to avoid surprising others.

Yes, that's maybe a bit too much effort.

bug#65622: Inappropriate suppression of backtrace on an error

[Alan Mackenzie]

Hello, Michael.

The bug is now fixed.

On Thu, Aug 31, 2023 at 01:16:42 +0200, Michael Heerdegen wrote:
> Alan Mackenzie <acm@muc.de> writes:

> > Hello, Emacs.
> > (v) Set debug-on-error to t with M-: (setq debug-on-error t).
> > (vi) Repeat (iv).  This throws the same error, without a backtrace.  This
> >   lack of a backtrace is a bug.

> I think you just need (setq eval-expression-debug-on-error t).

> Stumbled over the same issue ... yesterday?, or so.  Also when debugging
> Edebug.

> Maybe people would like debug-on-error -> t taking precedence over
> eval-expression-debug-on-error -> nil?  The danger is they turn both off
> by default and then forget about the second one.

> Michael.

-- 
Alan Mackenzie (Nuremberg, Germany).

bug#65622: Inappropriate suppression of backtrace on an error

[Michael Heerdegen]

Alan Mackenzie <acm@muc.de> writes:

> Hello, Michael.
>
> The bug is now fixed.

Great, thanks [...looking at the fix...] so this had nothing to do at
all with `eval-expression-debug-on-error'?

Michael.

bug#65622: Inappropriate suppression of backtrace on an error

[Alan Mackenzie]

Hello, Michael.

On Thu, Sep 21, 2023 at 03:44:47 +0200, Michael Heerdegen wrote:
> Alan Mackenzie <acm@muc.de> writes:

> > The bug is now fixed.

> Great, thanks [...looking at the fix...] so this had nothing to do at
> all with `eval-expression-debug-on-error'?

Well, not really.  But having eval-expression-debug-on-error set now
causes useful backtraces on the particular error.  In fact, two
successive backtraces, then a bare error message for a single error.
This is not ideal, but IMHO better than what came before.

And there was a devilish typo there, too, which took me days to find.
:-(

> Michael.

-- 
Alan Mackenzie (Nuremberg, Germany).